Improving Multi-Site Autism Classification via Site-Dependence Minimization and Second-Order Functional Connectivity.

Researchers developed better computer programs to identify autism using brain scans from multiple research centers. They created a new way to analyze how different brain regions connect and communicate. A major problem was that brain scans from different centers varied in quality and characteristics. By accounting for these differences, they improved the accuracy of autism identification to 73%, which is better than previous methods.

This study developed improved machine learning methods to classify autism spectrum disorder using brain imaging data from multiple research sites. The researchers used the ABIDE dataset and introduced a new method called 'Tangent Pearson embedding' to better analyze functional brain connectivity patterns. A key challenge they addressed was that different research sites produce varying data characteristics, which can reduce classification accuracy. By minimizing these site-dependent differences through domain adaptation techniques, they achieved 73% classification accuracy - an improvement over existing state-of-the-art methods.

The study found statistically significant relationships between research sites and brain connectivity features, highlighting the importance of accounting for multi-site data variations in autism classification models.

This technical advancement in neuroimaging analysis methods may eventually contribute to more accurate autism diagnostic tools. However, 73% accuracy suggests current limitations for direct clinical application. The approach addresses important challenges in combining data from multiple research sites.

Sample size not reported. Classification accuracy of 73% still indicates substantial misclassification. Study focuses on technical methodology rather than clinical validation. Generalizability to clinical settings unclear.

Machine learning has been widely used to develop classification models for autism spectrum disorder (ASD) using neuroimaging data. Recently, studies have shifted towards using large multi-site neuroimaging datasets to boost the clinical applicability and statistical power of results. However, the classification performance is hindered by the heterogeneous nature of agglomerative datasets. In this paper, we propose new methods for multi-site autism classification using the Autism Brain Imaging Data Exchange (ABIDE) dataset.

We firstly propose a new second-order measure of functional connectivity (FC) named as Tangent Pearson embedding to extract better features for classification. Then we assess the statistical dependence between acquisition sites and FC features, and take a domain adaptation approach to minimize the site dependence of FC features to improve classification. Our analysis shows that 1) statistical dependence between site and FC features is statistically significant at the 5% level, and 2) extracting second-order features from neuroimaging data and minimizing their site dependence can improve over state-of-the-art (SOTA) classification results, achieving a classification accuracy of 73%. The code is available at https://github.com/kundaMwiza/fMRI-site-adaptation.